A new approach for brain-computer interface (BCI) systems

Within the scope of this thesis, the proposed hybrid BCI system was tested using two different EEG devices and a comprehensive evaluation was made in terms of both technical performance and user experience. In the system implemented with the Emotiv Epoc X device, an innovative structure was presented in which Steady State Visually Evoked Potential (SSVEP) and EOG signals are evaluated together, reducing the visual stimulus dependency in traditional visual stimuli-based approaches. In order to reduce the negative effects of visual stimuli that require the user to focus directly, a single low-frequency (7Hz) LED was placed on top of the screen and this LED served as a control layer that verified voluntary activation. The user orientation was successfully determined by EOG artefacts obtained from EEG signals. In the two-stage classification, the Random Forest (RF) algorithm provided 99.57% accuracy in the first stage and 97.83% accuracy and 36.75 (bit/min) Information Transfer Rate (ITR) in the second stage. In the overall evaluation, the technical performance of the system was verified with 97.42% accuracy and 35.75 (bit/min) ITR. On the other hand, a similar two-stage hybrid structure was applied in the other system developed with the Emotiv Flex EEG device; SSVEP and EOG signals were evaluated sequentially. In the first stage, safe activation was achieved with the help of LED at 7Hz frequency, and in the second stage, four-way object tracking was performed with EOG signals. This system was designed with the aim of not only technical accuracy but also inter-session stability. The data obtained in two different sessions were processed with the Correlation Alignment (CORAL) based domain adaptation method and it was shown that the system works reliably without the need for recalibration. Bagging Algorithm provided the highest performance with 99.12% accuracy in the first stage and 94.29% accuracy, 38.35 (bit/min) ITR, 94.55% precision and 94.42% F1-score in the second stage. In both systems, user comfort is prioritized; low-frequency LED preferences, motion-based guidance structure and limited number of EEG channels minimize both visual fatigue and system complexity. In particular, in the Emotiv Epoc X system, involuntary activations based on EOG are controlled, while in the Emotiv Flex system, inter-session stability is ensured. As a result, the hybrid BCI system developed using two devices has proven to be a strong candidate for the transition from laboratory environments to daily life by offering high classification performance, system stability and reliability without neglecting user ergonomics.